In [1]:
%load_ext autoreload
%autoreload 2
In [2]:
import pickle
import itertools
import numpy as np
import pandas as pd
from sklearn.preprocessing import KBinsDiscretizer as kbd
import matplotlib
import matplotlib.pyplot as plt
import seaborn as sns
from modules.utils.general_utils.utilities import group_wise_binning
from modules.utils.general_utils.visualizers import visualize_neurons_function
In [3]:
def sns_styleset():
sns.set(context='paper', style='ticks', font='DejaVu Sans')
matplotlib.rcParams['figure.dpi'] = 100
matplotlib.rcParams['axes.linewidth'] = 1
matplotlib.rcParams['xtick.major.width'] = 1
matplotlib.rcParams['ytick.major.width'] = 1
matplotlib.rcParams['xtick.major.size'] = 3
matplotlib.rcParams['ytick.major.size'] = 3
matplotlib.rcParams['xtick.minor.size'] = 2
matplotlib.rcParams['ytick.minor.size'] = 2
matplotlib.rcParams['font.size'] = 13
matplotlib.rcParams['axes.titlesize'] = 13
matplotlib.rcParams['axes.labelsize'] = 13
matplotlib.rcParams['legend.fontsize'] = 13
matplotlib.rcParams['xtick.labelsize'] = 13
matplotlib.rcParams['ytick.labelsize'] = 13
sns_styleset()
cmap = matplotlib.cm.get_cmap('Paired')
1 - Load Globals¶
In [4]:
SNAPSHOTS = 10
TARGETS = [
'tar_sessions',
'tar_delta_sessions',
'tar_active_time',
'tar_session_time',
'tar_activity'
]
TARGETS_RMP = {
'tar_delta_sessions': 'Future Absence',
'tar_active_time': 'Future Active Time',
'tar_session_time': 'Future Session Time',
'tar_activity': 'Future Session Activity',
'tar_sessions': 'Future N° Sessions'
}
with open('results\\saved_objects\\mappers\\context.pkl', 'rb') as pickle_file:
CONTEXT_RMP = pickle.load(pickle_file)
CONTEXT_RMP = {value: key for key, value in CONTEXT_RMP.items()}
2 - Load Data Container¶
In [5]:
with (open('results\\saved_data_containers\\melchior.pkl', 'rb')) as container:
data_container = pickle.load(container)
In [6]:
# columns_to_drop = np.random.choice([i for i in range(96)], 86, replace=False)
# columns_to_keep = [column for column in [i for i in range(96)] if column not in columns_to_drop]
columns_to_keep = [20,28,34,41,44,50,54,56,57,72]
columns_to_drop = [column for column in [i for i in range(96)] if column not in columns_to_keep]
3 - Full Panel¶
In [8]:
for snapshot in range(SNAPSHOTS):
temporal_context = data_container['context'][snapshot]
temporal_embe = np.load(f'results\\saved_emb\\melchior_eng_emb_{snapshot}.npy')
temporal_embe = temporal_embe[~np.isnan(temporal_embe).any(axis=1)]
df = pd.DataFrame(temporal_embe)
df['context'] = temporal_context
for target in TARGETS:
df[target] = data_container['prediction_ds'][target][snapshot]
df = df[df['context'] == 3]
columns = list(df.drop(TARGETS+['context'], axis=1))
df[columns] = kbd(
n_bins=100,
encode='ordinal',
).fit_transform(df[columns].values)
for target in TARGETS:
color = group_wise_binning(
array=df[target].values,
n_bins=100,
grouper=df['context'].values,
method='discret',
)
df[target] = color
melted = pd.melt(
df,
id_vars=TARGETS + ['context'],
var_name='Artificial Neurons',
value_name='Signal'
)
index = 0
fig = plt.figure(figsize=(9, 6), constrained_layout=True)
spec = fig.add_gridspec(ncols=6, nrows=2)
axs_metrics = [
fig.add_subplot(spec[0, 0:2]),
fig.add_subplot(spec[0, 2:4]),
fig.add_subplot(spec[0, 4:]),
fig.add_subplot(spec[1, 1:3]),
fig.add_subplot(spec[1, 3:5])
]
for metric, ax_metric in zip(TARGETS, axs_metrics):
grouped = melted.groupby(['Signal', 'Artificial Neurons'])[metric].mean()
grouped = grouped.reset_index().pivot('Artificial Neurons', 'Signal', metric)
grouped = grouped.transpose()
grouped = grouped.sort_values('Signal', ascending=True).values
grouped = np.delete(grouped,np.s_[columns_to_drop],axis=1)
img = ax_metric.imshow(
grouped,
cmap='coolwarm',
vmin=0,
vmax=100,
aspect='auto'
)
ax_metric.set_ylim(ax_metric.get_ylim()[::-1])
if index == 0:
ax_metric.set_xticks([])
elif index == 4:
ax_metric.set_yticks([])
ax_metric.set_xticks([tick for tick in range(len(columns_to_keep))])
ax_metric.set_xticklabels(columns_to_keep)
elif index == 3:
ax_metric.set_xticks([tick for tick in range(len(columns_to_keep))])
ax_metric.set_xticklabels(columns_to_keep)
else:
ax_metric.set_xticks([])
ax_metric.set_yticks([])
ax_metric.set_title(f'{TARGETS_RMP[metric]} - $t$ {snapshot +1}')
index += 1
fig.text(0.5, -0.01, 'Artificial Neurons', ha='center')
fig.text(-0.01, 0.5, 'Discretized Activation', va='center', rotation='vertical')
plt.tight_layout()
cbaxes = fig.add_axes([1.0, 0.12, 0.02, 0.80])
cbar = fig.colorbar(
img,
cax=cbaxes,
cmap='coolwarm',
boundaries=np.linspace(0, 100, 100),
ticks=[0, 25, 50, 75, 100]
)
cbar.set_label('Discretized Metric Value')
plt.savefig(
f'results\\figures\\embeddings\\activation_full_panel\\snaphot_{snapshot}',
dpi=500,
bbox_inches='tight'
)
plt.show()
4 - Temporal Panel¶
In [9]:
for metric in TARGETS:
fig = plt.figure(figsize=(9, 3), constrained_layout=True)
spec = fig.add_gridspec(ncols=3, nrows=1)
for column, snapshot in enumerate([1, 2, 3]):
temporal_context = data_container['context'][snapshot]
temporal_embe = np.load(f'results\\saved_emb\\melchior_eng_emb_{snapshot}.npy')
temporal_embe = temporal_embe[~np.isnan(temporal_embe).any(axis=1)]
df = pd.DataFrame(temporal_embe)
df['context'] = temporal_context
df[metric] = data_container['prediction_ds'][metric][snapshot]
df = df[df['context'] == 3]
columns = list(df.drop([metric, 'context'], axis=1))
color = group_wise_binning(
array=df[metric].values,
n_bins=100,
grouper=df['context'].values,
method='discret'
)
df[metric] = color
columns = list(df.drop(['context', metric], axis=1))
df[columns] = kbd(
n_bins=100,
encode='ordinal',
).fit_transform(df[columns].values)
melted = pd.melt(
df,
id_vars=[metric, 'context'],
var_name='Artificial Neurons',
value_name='Signal'
)
grouped = melted.groupby(['Signal', 'Artificial Neurons'])[metric].mean()
grouped = grouped.reset_index().pivot('Artificial Neurons', 'Signal', metric)
grouped = grouped.transpose()
grouped = grouped.sort_values('Signal', ascending=True).values
grouped = np.delete(grouped,np.s_[columns_to_drop],axis=1)
ax_metric = fig.add_subplot(spec[0, column])
img = ax_metric.imshow(
grouped,
cmap='coolwarm',
vmin=0,
vmax=100,
aspect='auto'
)
ax_metric.set_ylim(ax_metric.get_ylim()[::-1])
if column > 0:
ax_metric.set_yticks([])
ax_metric.set_xticks([tick for tick in range(len(columns_to_keep))])
ax_metric.set_xticklabels(columns_to_keep)
ax_metric.set_title(f'{TARGETS_RMP[metric]} - $t$ {snapshot+1}')
fig.text(0.5, -0.01, 'Artificial Neurons', ha='center')
fig.text(-0.01, 0.5, 'Discretized Activation', va='center', rotation='vertical')
plt.tight_layout()
cbaxes = fig.add_axes([1.0, 0.12, 0.02, 0.80])
cbar = fig.colorbar(
img,
cax=cbaxes,
cmap='coolwarm',
boundaries=np.linspace(0, 100, 100),
ticks=[0, 25, 50, 75, 100]
)
cbar.set_label('Discretized Metric Value')
plt.savefig(
f'results\\figures\\embeddings\\activation_temporal_panel\\{TARGETS_RMP[metric]}',
dpi=500,
bbox_inches='tight'
)
plt.show()
plt.close()
5 - Neurons Functions¶
In [10]:
for metric in TARGETS:
for column, snapshot in enumerate([0, 1, 2, 3]):
temporal_context = data_container['context'][snapshot]
temporal_embe = np.load(f'results\\saved_emb\\melchior_eng_emb_{snapshot}.npy')
temporal_embe = temporal_embe[~np.isnan(temporal_embe).any(axis=1)]
df = pd.DataFrame(temporal_embe)
df['context'] = temporal_context
df[metric] = data_container['prediction_ds'][metric][snapshot]
df = df[df['context'] == 3]
columns = list(df.drop([metric, 'context'], axis=1))
color = group_wise_binning(
array=df[metric].values,
n_bins=100,
grouper=df['context'].values,
method='discret'
)
df[metric] = color
columns = list(df.drop(['context', metric], axis=1))
df[columns] = kbd(
n_bins=100,
encode='ordinal',
).fit_transform(df[columns].values)
melted = pd.melt(
df,
id_vars=[metric, 'context'],
var_name='Artificial Neurons',
value_name='Signal'
)
visualize_neurons_function(
data=melted,
metric=metric,
snapshot=snapshot+1,
neurons=columns_to_keep,
metric_rmp=TARGETS_RMP
)
